The three mainstream writing approaches for Kling 3.0 (Kuaishou): 4-part basic / 5-layer advanced / image-to-video.
When not to use this skill
- Complex multi-shot storyboard narrative (>15s) → use seedance-storyboard
- 5/10/15-second compact short films with low demand for native audio → use happyhorse-prompter
- Existing Seedance prompts that need fixing → use seedance-debugger
Decision tree for the three writing approaches
你的需求是？
├── 5 秒短视频、单一场景 → 写法 1（4 部分基础公式）
├── 带剧情、带音频、多角色 → 写法 2（5 层进阶公式）★ Kling 强项
└── 已有参考图，只需动起来 → 写法 3（图生视频专用）★ Kling 强项
Workflow
Step 1: Determine the writing approach
| User's need keywords | Writing approach |
|---|---|
| Short videos, TikTok, Douyin, loops, viral hits, transformations | Approach 1 (4-part) |
| Dialogue, narrative, voiceover, CTA, ads, multiple characters | Approach 2 (5-layer) |
| This image, refer to this photo, make it move | Approach 3 (image-to-video) |
Step 2: Clarify missing elements (when necessary)
If there isn't enough information, ask 2-3 key questions (not all of them):
| Scenario | The 2-3 must-ask items |
|---|---|
| Short video | Subject (specific appearance), action (verb + endpoint), style (cinematic/anime/cyberpunk) |
| 5-layer narrative | Character (age + clothing), dialogue content, whether native audio is needed (music/sound effects/voiceover) |
| Image-to-video | What change the subject undergoes in the image (do not describe the image itself again), how the camera moves |
Step 3: Fill in according to the chosen writing approach
Approach 1: 4-part basic formula
[Subject - 具体视觉细节] [Action - 精确运动+终点] [Context - 3-5 环境元素] [Style - 镜头+灯光+氛围].
Detailed reference methodology/09-kling-formulas.md。
Approach 2: 5-layer advanced formula (Kling's signature strength)
Strongly recommend using a structured format, the model parses by layer:
Scene: [地点 + 时间 + 氛围]
Characters: [每个角色的具体描述，age + 关键服装/外貌]
Action: [按时间顺序的步骤，体现因果与递进]
Camera: [镜头类型 → 运动 → 切换]
Audio & Style: [对话/音效/音乐 + 视觉风格]
⚠️ The Audio & Style layer is Kling's differentiating selling point, be sure to write it! Seedance / HappyHorse cannot achieve character-directed speech.
Approach 3: Dedicated to image-to-video
Iron rule: describe motion only, do not repeat visual information already in the image。
❌ Do not write:A red sports car with chrome wheels on a wet road...(already in the image)
✅ Write directly:The headlights blaze on and the engine roars to life. The rear tires spin...
Step 4: Self-check against the 6 Kling writing guidelines
- Use active verbs to replace vague words (rockets / slams / erupts)
- Describe the temporal flow(beginning → middle → end)
- Use specific light sources instead of "dramatic lighting"
- Add tactile details（grain / reflections / sweat / steam）
- Add an explicit endpoint(prevents 99% of freezes):
then settles back / fades out - For fantasy scenes, use physical interaction anchoring（stone crumbles / torches flicker）
Step 5: Decide the language (Chinese or English)
| Scenario | Recommended language |
|---|---|
| Chinese aesthetics, wuxia, hanfu, Chinese dialogue | Chinese(Kling has the strongest Chinese understanding) |
| International, Hollywood style, product ads | English |
| Mix of both | Choose a dominant style,don't mix |
Step 6: Add negative prompts
Negative: cartoonish, 3D render, smooth plastic skin, floating limbs,
sliding feet, text morphing, no watermark, no logo.
When you need a real-person feel, add smiling to negative (to make expressions natural).
Step 7: Output format
## 生成的 Kling 提示词（写法 N）
\`\`\`
[完整提示词]
\`\`\`
**推荐参数**：宽高比 [X:Y] · 时长 [N 秒] · 模型 [Kling 3.0 / Kling O1]
**关键技巧**：[1-2 句说明这条提示词的设计思路]
**可调点**：[告诉用户可以怎么改]
Practical examples
Example 1: The user wants "a short video, a girl turns into an anime character"
Judgment: transformation viral hit → Writing method 1 (4-part basic)
Output：
A young woman in casual streetwear standing center frame, her body slowly engulfed by glowing particles as she transforms into a vibrant anime version of herself with stylized hair and large expressive eyes, then settles into her new form. TikTok-style vertical framing, neon pink and cyan rim lighting, smooth transition with sparkle bursts, high-energy K-pop MV aesthetic.
Negative: 3D render, plastic skin, floating limbs, text morphing, watermark.
Parameters：9:16 · 5s · Kling 3.0
Key techniques: All 4 elements present — Subject (specific clothing) + Action (with an endpoint:then settles) + Context (3: neon rim light/sparkle/vertical) + Style (K-pop MV aesthetic).
Adjustable points: Replace K-pop MV with Cyberpunk / Studio Ghibli to change the style; replace glowing particles with glitch effects to change the transformation mechanism.
Example 2: The user wants "two old friends reuniting after 10 years, with voiceover"
Judgment: With plot + with audio → Writing method 2 (5-layer advanced)
Output：
Scene: A park bench at golden hour in late autumn, fall leaves drifting through the air, soft amber sunlight slanting through bare branches.
Characters: Two reunited friends with a 10-year gap, both early 30s. Character A wears a wool coat and scarf, slightly nervous. Character B wears a denim jacket, more outwardly emotional.
Action: Character A sits down beside Character B, hesitates, then says "It's been a long time." Character B turns slowly, eyes widening with recognition, replies "Too long. I almost didn't recognize you." They both pause, then share a quiet laugh that breaks the tension.
Camera: Slow push-in on Character A as he speaks, then over-the-shoulder cut to Character B's reaction, ending on a held wide two-shot of both on the bench.
Audio & Style: Soft solo piano underscore building gently, ambient park sounds (distant kids playing, leaves rustling), warm cinematic color grade with teal-and-amber accents, naturalistic dialogue delivery.
Negative: smiling broadly, cartoonish, 3D render, plastic skin, sliding feet, watermark.
Parameters: 16:9 · 12s · Kling 3.0 (recommended)
Key techniques: All 5 layers enabled. The Characters layer gives each of the two characters their own "anchor clothing" to prevent drift; the Action layer uses a causal chain (sit down → hesitate → speak → react → laugh together) to make the narrative natural; the Audio & Style layer provides both music and ambient sound, triggering Kling's native audio-visual sync.
Adjustable points: Change golden hour → blue hour to change the tone; change piano underscore → silence + only ambient to make the emotion more restrained.
Example 3: The user uploaded a kitchen photo and wants to "make it move, with a chef tossing and stir-frying"
Judgment: Image-to-video → Writing method 3
Output：
The chef flips the wok hard, launching noodles into the air as a fireball erupts upward. He catches them clean, tilts the wok and slides the dish onto a plate with his spatula, then pushes the plate across the worn counter toward the camera with a grin. Steam billows upward through the lantern light.
Negative: floating limbs, plastic skin, sliding feet, watermark.
Parameters：16:9 · 8s · Kling 3.0
Key techniques: Do not re-describe the kitchen layout/chef's appearance in the image at all (the image already has them); only write the motion sequence (flip → fire → catch → slide → push) + physical details (steam billows). Open flames creating dynamic lighting is a strength of Kling.
Adjustable points: Delete fireball erupts to change to gentle stir-frying; delete with a grin to make the expression more focused.
Reference resources
- Kling complete methodology:methodology/09-kling-公式.md
- 36 tested prompts:prompts/kling/README.md
- Camera movement dictionary (general):methodology/05-运镜词典.md
- Official platform:klingai.com